# LucyWorks

> AI training for enterprise procurement teams. Twenty short lessons that each
> take one real procurement workflow — contract review, quote normalization,
> savings reports, supplier risk — and work it end to end with AI (ChatGPT,
> Claude, Gemini, Copilot), with real platform evidence and practice case files.
> The course is a one-time purchase; teams can book a live workshop.

## Core pages
- [The 20 Lessons](https://www.lucyworks.com/lessons): the full course catalogue; lesson 2 is free to read.
- [Answers](https://www.lucyworks.com/answers): direct, evidence-backed answers to "can AI do X in procurement?" plus a procurement-AI glossary.
- [PAIR-20](https://www.lucyworks.com/pair-20): the open procurement-AI benchmark — real dated runs, a fabricated fact fails the task.
- [Reality Check](https://www.lucyworks.com/reality-check): monthly field notes — the same procurement task on four AI platforms, misses included.
- [Pricing](https://www.lucyworks.com/pricing): €199 one-time for individuals, €2,500 live workshop for teams.
- [Book a call](https://www.lucyworks.com/book): schedule the team workshop.
- [About](https://www.lucyworks.com/about): who runs LucyWorks and why the course exists.

## Lessons
- [Normalize Messy SaaS Vendor Quotes — Unit Conversion, Hidden Fees, Cross-Platform Verification](https://www.lucyworks.com/lessons/saas-quote-normalization): Shows how to feed AI three messy, differently-formatted SaaS vendor quotes (plus a distractor) so it normalizes them onto the same basis instead of comparing sticker prices that hide a per-seat trap.
- [Review a SaaS Master Service Agreement for Traps — Auto-Renewal, Price Escalation, Data Portability](https://www.lucyworks.com/lessons/saas-msa-review): Walks through reviewing a SaaS master service agreement with AI to catch an auto-renewal notice trap, a price-escalation clause disguised as a cap, and a data-portability exit cost — then turning those findings into a legal question list, a VP briefing, and a vendor counter-proposal.
- [Compress a Messy Savings Ledger Into the One Page Your CFO Will Sign](https://www.lucyworks.com/lessons/cfo-savings-report): Teaches how to take a messy, mixed savings ledger and get AI to correctly separate hard cost reductions from cost avoidance, then compress the audited numbers into a one-page report a CFO will actually sign.
- [The Price-Increase Counter: Make AI Show You What It Can't Know](https://www.lucyworks.com/lessons/commodity-price-counter): Shows how to get AI to admit it cannot fetch live commodity index data, supply the missing external price data yourself, and use that to build a fact-based counter to a supplier's price-increase request.
- [SaaS License Audit — Checking Contracts Before Cutting Seats](https://www.lucyworks.com/lessons/saas-license-audit): Covers auditing SaaS license usage against contract terms to find unused seats worth reclaiming, while avoiding true-up and minimum-commitment traps hidden in the contract language.
- [Build the AI-Adoption ROI Case That Survives the CFO](https://www.lucyworks.com/lessons/ai-adoption-roi): Builds a defensible AI-adoption ROI case that accounts for the hidden review-tax cost of checking AI output, structured to survive CFO scrutiny rather than overstating savings.
- [Compare Three Consulting Bids With AI — The $15K Spread That's Really $750](https://www.lucyworks.com/lessons/consulting-rate-comparison): Compares three consulting bids with different rate structures and expense terms so the true cost spread is visible, rather than the misleading headline-rate comparison.
- [Commodity-Price Drift Monitor — Make the AI Design the Checklist, Then Bring the Index](https://www.lucyworks.com/lessons/commodity-price-monitor): Has AI design an ongoing commodity price monitor built around index data it cannot fetch itself, so the buyer knows exactly what data to feed it each period and how to read the resulting trend.
- [Auditing a Services SOW Bundle: Feed, Chain, Report](https://www.lucyworks.com/lessons/services-sow-audit): Shows how to feed AI a full bundle of related documents — a statement of work with leaked tracked-change comments, a rate card, the governing MSA, and a prior closed SOW as a distractor — so it catches protections quietly removed rather than reviewing the SOW in isolation.
- [PO Price-Variance Audit — From Raw Extract to Three Director Actions](https://www.lucyworks.com/lessons/po-price-variance): Teaches AI to check purchase order pricing against the actual contracted rules (not just against the numbers on the PO), catching variances a numbers-only comparison would miss.
- [Freight Landed Cost — Normalizing Quotes That Cover Different Spans of the Same Trip](https://www.lucyworks.com/lessons/freight-landed-cost): Compares three freight quotes on true landed cost — surcharges, fees, and currency terms included — so the quote that looks cheapest on the headline rate isn't wrongly chosen.
- [Make AI Cross-Reference Three Documents Before You Renew a Supplier](https://www.lucyworks.com/lessons/supplier-financial-health): Shows how to have AI cross-reference three separate documents (financials, governance disclosures, debt schedule) before a supplier renewal decision, catching a risk pattern that only appears when the documents are read together.
- [SaaS Consolidation: From Three Overlapping Tools to One Committee-Ready Memo](https://www.lucyworks.com/lessons/saas-consolidation): Turns a list of overlapping SaaS tools into a consolidation analysis — identifying redundant spend and building the case for which tools to cut or merge.
- [Build a Supplier Risk Register When Three of Your Data Fields Are Lying](https://www.lucyworks.com/lessons/supplier-risk-register): Builds a supplier risk register while working around a data-integrity problem: several fields in the source ERP and supplier-master data are simply wrong, and the lesson teaches how to catch and correct them before scoring risk.
- [Compare Two Freight Contracts by Making AI Build the Crosswalk](https://www.lucyworks.com/lessons/freight-sla-comparison): Compares two freight contracts by having AI build a clause-by-clause crosswalk, surfacing how definitions like on-time delivery and force majeure interact to change the effective service level.
- [Turn Three Fleet Bids Into One Recommendation Your VP Can Sign](https://www.lucyworks.com/lessons/fleet-bid-scorecard): Turns three competing fleet bids into a single weighted scorecard and recommendation that a VP can approve, instead of an apples-to-oranges bid comparison.
- [Build a Weighted Vendor Ranking from Three Data Sources](https://www.lucyworks.com/lessons/vendor-performance-summary): Combines three scattered vendor performance data sources into one ranking a director can sign off on, resolving conflicts between the sources along the way.
- [From Three Incompatible Equipment Proposals to One Number Your Capital Committee Can Approve](https://www.lucyworks.com/lessons/lease-vs-buy): Compares three incompatible equipment lease/buy proposals by normalizing hidden costs (like wear items and residual value) into one comparable number a capital committee can approve.
- [Merge Three Catalogs Without Deleting the One Part Someone's Safety Depends On](https://www.lucyworks.com/lessons/mro-catalog-rationalization): Merges three overlapping MRO parts catalogs while catching false duplicates and specification gaps, so a safety-critical part doesn't get deleted by mistake during consolidation.
- [Turn Three Maintenance Bids Into One Five-Year Cost of Ownership](https://www.lucyworks.com/lessons/facilities-maintenance-tco): Turns three facilities-maintenance bids into a five-year total cost of ownership comparison and a one-page VP decision brief, rather than comparing headline service fees.

## Answers
- [Can AI build a supplier risk register?](https://www.lucyworks.com/answers/can-ai-build-a-supplier-risk-register): Yes, but it still needs field-level verification. In PAIR-20's July 2026 T04 run, most platforms flagged hidden source-status risks, while Copilot either buried Weiss or fabricated a Dravec expiry.
- [Can AI calculate landed cost correctly?](https://www.lucyworks.com/answers/can-ai-calculate-landed-cost): It can build the landed-cost stack, but do not trust the final number. In PAIR-20's July 2026 run, all four platforms treated the headline freight quote as incomplete, and none matched the EUR7,940 case answer.
- [Can AI catch a unit-of-measure error in a quote comparison?](https://www.lucyworks.com/answers/can-ai-catch-a-unit-of-measure-error): Sometimes — and you cannot predict which platform will trip. In PAIR-20's July 2026 run, Gemini caught a per-100-seats unit trap unprompted; Claude carried the wrong number into its ranking even after being told the unit.
- [Can AI check a contract before your meeting?](https://www.lucyworks.com/answers/can-ai-check-a-contract-before-a-meeting): Yes, for a first pass. In July 2026, all four platforms found the renewal clauses, but three needed direction before connecting the 60-day price notice to the 90-day cancellation deadline.
- [Can AI compare supplier quotes reliably?](https://www.lucyworks.com/answers/can-ai-compare-supplier-quotes): Not reliably. In PAIR-20's July 2026 run, AI found several quote issues, but one platform carried a unit error into the ranking and another ranked the highest consulting bid first after doing the math.
- [Can AI find hidden fees in a quote?](https://www.lucyworks.com/answers/can-ai-find-hidden-fees-in-a-quote): Yes, sometimes. In PAIR-20's July 2026 run, platforms found the SaaS mandatory fee and surfaced consulting hidden-cost items, but later failures still changed the recommendation and blended-rate presentation.
- [Can AI help prepare a supplier negotiation?](https://www.lucyworks.com/answers/ai-procurement-negotiation): Yes — AI can draft position papers, comparison tables, and clause lists. In PAIR-20's July 2026 run, the platforms surfaced clauses and built cost stacks; negotiation judgment stayed with the buyer.
- [Can AI merge and deduplicate a parts catalog?](https://www.lucyworks.com/answers/can-ai-merge-a-parts-catalog): Yes, with a narrow review step. In PAIR-20's July 2026 T06 run, most platforms separated the glove false duplicate, but Copilot filled Atlas's blank box quantity with "Likely 100ct."
- [Can AI rank bids correctly once the math is done?](https://www.lucyworks.com/answers/can-ai-rank-bids-correctly): Sometimes. In PAIR-20's July 2026 T05.5 run, ChatGPT calculated the consulting-bid math correctly, then ranked the highest all-in total first in its summary table.
- [Can AI review a procurement contract?](https://www.lucyworks.com/answers/can-ai-review-a-procurement-contract): Yes, but a clause list is not enough. In PAIR-20's July 2026 run, all four platforms found the renewal clauses; three needed direction before connecting the 60-day price notice to the 90-day cancellation deadline.
- [Can AI tell true duplicates from false twins in item data?](https://www.lucyworks.com/answers/can-ai-spot-a-false-duplicate): Sometimes, but only if the model keeps blank fields visible. In PAIR-20's July 2026 T06 run, ChatGPT, Claude, and Gemini flagged the glove false twin; Copilot guessed a missing box count.
- [Can you trust AI's arithmetic in procurement files?](https://www.lucyworks.com/answers/can-you-trust-ai-arithmetic-in-procurement): Not as the final check. In PAIR-20's July 2026 T03 run, four platforms rebuilt the landed-cost stack, but every final number differed from the EUR7,940 case answer.
- [Can you use Microsoft Copilot for procurement work?](https://www.lucyworks.com/answers/copilot-for-procurement): Yes, on the free or standard tier. In a July 2026 test, Copilot made arithmetic and fabrication errors on real procurement tasks — treat its output as a draft to check, not an answer.
- [Do procurement teams already use AI?](https://www.lucyworks.com/answers/do-procurement-teams-already-use-ai): Yes. Microsoft Work Trend Index 2024 says 75% of knowledge workers use AI and 78% of those AI users bring their own tools, while only 39% have received company AI training.
- [Does a better prompt fix AI's procurement mistakes?](https://www.lucyworks.com/answers/does-a-better-prompt-fix-ai-mistakes): No. In PAIR-20's July 2026 run, structured briefs made outputs easier to audit, but did not prevent a unit miss, wrong landed-cost numbers, a ranking contradiction, or stripped hidden costs.
- [Does generic AI training work for procurement teams?](https://www.lucyworks.com/answers/does-generic-ai-training-work): Generic AI training is a weak bet by itself. Section AI Report 2026 found no measurable productivity gain; CIPD says only 7% of organizations have a learning-transfer process.
- [How can AI actually help procurement?](https://www.lucyworks.com/answers/how-can-ai-help-procurement): It reads messy files, builds comparison tables, and flags obvious problems — a useful first pass. It gets math wrong, misses how clauses affect each other, and guesses when data is missing, so always check the output.
- [How can AI help counter a supplier price increase?](https://www.lucyworks.com/answers/how-can-ai-counter-a-supplier-price-increase): By rebuilding the cost stack line by line. In PAIR-20's July 2026 run, all four platforms broke a freight quote into its landed-cost layers — giving a buyer the structure to challenge the increase, not just the headline number.
- [How do you use AI in procurement?](https://www.lucyworks.com/answers/how-to-use-ai-in-procurement): Treat it as a drafting tool, not an oracle: give it a clear brief, let it organize the work, then check every number and clause yourself before it shapes a decision.
- [How do you use ChatGPT in procurement?](https://www.lucyworks.com/answers/chatgpt-for-procurement): Use it to organize procurement files and draft a first pass — not to trust the final numbers. In a July 2026 comparison test, ChatGPT read contracts well but made arithmetic errors. Check its totals yourself.
- [How often does AI fabricate facts in procurement files?](https://www.lucyworks.com/answers/how-often-does-ai-fabricate-in-procurement): PAIR-20 does not give a stable frequency. In one July 2026 set of six single-run tasks, two baited tasks recorded fabricated facts: an invented Dravec expiry and a guessed Atlas box quantity.
- [Is ChatGPT safe for procurement work?](https://www.lucyworks.com/answers/is-chatgpt-safe-for-procurement): It is safe if you control what goes in. The risk is pasting real supplier prices, terms, or volumes into a public-tier tool — not the tool itself. Treat input as the security boundary.
- [Is free-tier AI good enough for procurement work?](https://www.lucyworks.com/answers/is-free-tier-ai-good-enough-for-procurement): Enough for triage, not enough for unsupervised decisions. In PAIR-20's July 2026 run, all findings came from free or standard web tiers, which produced useful drafts and still made arithmetic, ranking, and fabrication errors.
- [Is there a benchmark of AI performance on procurement tasks?](https://www.lucyworks.com/answers/is-there-a-benchmark-of-ai-performance-on-procurement-tasks): PAIR-20 is the closest published reference: six synthetic procurement tasks, four platforms, one dated July 2026 run. It shows where models agree and disagree, not a ranking—one run cannot be a leaderboard.
- [What are the best ChatGPT prompts for procurement?](https://www.lucyworks.com/answers/best-chatgpt-prompts-for-procurement): A better prompt makes AI output easier to check, not more accurate. In our July 2026 test, sharper prompts still let wrong numbers through. Build a clear brief and verify the numbers yourself.
- [What are the real use cases for AI in procurement?](https://www.lucyworks.com/answers/ai-procurement-use-cases): The tested use cases are paperwork-heavy first drafts: comparing quotes, reviewing contracts, rebuilding cost breakdowns, drafting risk registers, scoring bids, and cleaning up catalogs. Each produced a useful draft, but none was safe to use without a human check.
- [What belongs on an AI procurement checklist?](https://www.lucyworks.com/answers/ai-procurement-checklist): At minimum: a structured brief, unit/date/blank-field verification, a two-platform divergence check, and a no-unverified-numbers rule. Each item traces to a recorded failure in the July 2026 PAIR-20 run.
- [What do Incoterms mean for a landed-cost comparison?](https://www.lucyworks.com/answers/what-is-incoterms-scope): Incoterms define which parts of the shipment each quote covers. In landed-cost comparison, they decide which freight, insurance, handling, customs, and surcharge layers must be added before prices are comparable.
- [What does AI training for procurement teams actually look like?](https://www.lucyworks.com/answers/ai-training-for-procurement): It teaches your team to run real procurement workflows — supplier research, contract review, spend analysis — with AI tools they already have. Not a theory course. Each person builds something they'll use next week.
- [What is a blended rate in consulting bids?](https://www.lucyworks.com/answers/what-is-a-blended-rate): A blended rate is one average hourly rate for a consulting team, usually total fees divided by total hours. It hides the role mix, so procurement should check who does the work.
- [What is a freight SLA?](https://www.lucyworks.com/answers/what-is-a-freight-sla): A freight SLA is the service-level agreement inside a freight contract: the measurable delivery, damage, reporting, credit, and exception rules that decide whether carrier performance is acceptable.
- [What is a statement-of-work audit?](https://www.lucyworks.com/answers/what-is-a-sow-audit): A statement-of-work audit checks whether an SOW's scope, fees, timelines, acceptance terms, change control, and governing agreement match the buyer's intent before signature.
- [What is a structured brief (vs. a prompt)?](https://www.lucyworks.com/answers/what-is-a-structured-brief): A structured brief is a reusable instruction set: role, fields to check, rules, output format, and verification steps. A prompt is often a one-off request. The brief makes the work easier to audit.
- [What is a supplier risk register?](https://www.lucyworks.com/answers/what-is-a-supplier-risk-register): A supplier risk register is a ranked table of suppliers, the risks each one carries, the evidence behind those risks, and the actions needed to reduce exposure or prepare a fallback.
- [What is AI fabrication in procurement documents?](https://www.lucyworks.com/answers/what-is-ai-fabrication): AI fabrication is a fact the model adds without support in the source file. In procurement, that can mean an invented expiry date, assumed pack quantity, or any unsupported field in a deliverable.
- [What is an MSA (Master Service Agreement)?](https://www.lucyworks.com/answers/what-is-an-msa): An MSA, or Master Service Agreement, is the base contract that sets the legal and commercial rules for future orders, renewals, data rights, liability, termination, and related work.
- [What is BYOAI (bring your own AI)?](https://www.lucyworks.com/answers/what-is-byoai): BYOAI means employees use AI tools they choose themselves at work. Microsoft Work Trend Index 2024 says 78% of AI users bring their own AI tools to work.
- [What is generative AI for procurement, and how do you actually learn it?](https://www.lucyworks.com/answers/generative-ai-for-procurement): Generative AI tools — ChatGPT, Claude, Copilot, Gemini — produce new text and analysis from a prompt. In procurement, that means drafting, summarizing, and analyzing RFPs, contracts, and supplier communications on demand, not just retrieving stored data.
- [What is landed cost?](https://www.lucyworks.com/answers/what-is-landed-cost): Landed cost is the all-in buyer cost after adding the price, freight, duties, insurance, handling, currency effects, and other charges needed to get goods to the agreed destination.
- [What is learning transfer in AI training?](https://www.lucyworks.com/answers/what-is-learning-transfer): Learning transfer is the process that makes training show up in daily work. For AI training, it means people use the method on real tasks after the session ends.
- [What is maverick spend?](https://www.lucyworks.com/answers/what-is-maverick-spend): Maverick spend is purchasing that bypasses approved contracts, preferred suppliers, or procurement processes — whether by choice, habit, or missing coverage. It erodes negotiated prices and makes contract leakage harder to find.
- [What is PO price variance?](https://www.lucyworks.com/answers/what-is-po-price-variance): PO price variance is the difference between the price on a purchase order and the price the buyer should have paid under the contract, after unit, freight, discount, and escalation rules are applied.
- [What is quote normalization in procurement?](https://www.lucyworks.com/answers/what-is-quote-normalization): Quote normalization means converting supplier offers onto the same basis before ranking them: same unit, volume, term, fees, and exclusions. In July 2026, PAIR showed a per-100-seat unit could still trip AI.
- [What is savings realization (reported vs. realized)?](https://www.lucyworks.com/answers/what-is-savings-realization): Savings realization is the test of whether a reported procurement saving actually becomes lower spend, avoided cost, or another verified financial outcome after the deal is signed.
- [What is spend analysis (and can AI do it)?](https://www.lucyworks.com/answers/what-is-spend-analysis): Spend analysis is the process of collecting, cleansing, classifying, and examining purchase data to find savings. AI can draft classifications; for the totals, PAIR-20's July 2026 runs showed every platform's arithmetic drifted from case answers.
- [What is tail spend?](https://www.lucyworks.com/answers/what-is-tail-spend): Tail spend is the high-volume, low-value portion of a procurement budget — many suppliers and transactions but a small share of total spend — often unmanaged because no single purchase justifies a sourcing project.
- [What is total cost of ownership (TCO)?](https://www.lucyworks.com/answers/what-is-total-cost-of-ownership): Total cost of ownership is the full cost of buying, operating, maintaining, and exiting an asset or service, not just the quoted price or annual fee.
- [Which AI platform is best for procurement work?](https://www.lucyworks.com/answers/which-ai-platform-is-best-for-procurement): There is no single winner from the July 2026 PAIR-20 run. ChatGPT, Claude, Gemini, and Copilot each passed some tasks and missed others; the useful signal was where they disagreed.
- [Will AI invent numbers in procurement work?](https://www.lucyworks.com/answers/will-ai-invent-numbers-in-procurement): Yes. In the July 2026 PAIR run, Copilot fabricated a backup-contract expiry in a risk register and guessed an unstated box quantity in a catalog merge. Under the rubric, each fabrication failed the task.